[Bugfix][VLM] Add fallback to SDPA for ViT model running on CPU backend

[Isotr0py]

FILL IN THE PR DESCRIPTION HERE

FIX #xxxx (link existing issues this PR will resolve)

• [VLM][Model] TP support for ViTs #7186 enforced xformers on encoder attention for each ViT model, which broke the VLM support for CPU backend.
• This PR aims to add fallback to SDPA for CPU backend to fix this.

TODO

• Blip
• Clip
• InternViT
• Siglip

FIX #8097
BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE

---

PR Checklist (Click to Expand)

Thank you for your contribution to vLLM! Before submitting the pull request, please ensure the PR meets the following criteria. This helps vLLM maintain the code quality and improve the efficiency of the review process.

PR Title and Classification

Only specific types of PRs will be reviewed. The PR title is prefixed appropriately to indicate the type of change. Please use one of the following:

• [Bugfix] for bug fixes.
• [CI/Build] for build or continuous integration improvements.
• [Doc] for documentation fixes and improvements.
• [Model] for adding a new model or improving an existing model. Model name should appear in the title.
• [Frontend] For changes on the vLLM frontend (e.g., OpenAI API server, LLM class, etc.)
• [Kernel] for changes affecting CUDA kernels or other compute kernels.
• [Core] for changes in the core vLLM logic (e.g., LLMEngine, AsyncLLMEngine, Scheduler, etc.)
• [Hardware][Vendor] for hardware-specific changes. Vendor name should appear in the prefix (e.g., [Hardware][AMD]).
• [Misc] for PRs that do not fit the above categories. Please use this sparingly.

Note: If the PR spans more than one category, please include all relevant prefixes.

Code Quality

The PR need to meet the following code quality standards:

• We adhere to Google Python style guide and Google C++ style guide.
• Pass all linter checks. Please use format.sh to format your code.
• The code need to be well-documented to ensure future contributors can easily understand the code.
• Include sufficient tests to ensure the project to stay correct and robust. This includes both unit tests and integration tests.
• Please add documentation to docs/source/ if the PR modifies the user-facing behaviors of vLLM. It helps vLLM user understand and utilize the new features or changes.

Notes for Large Changes

Please keep the changes as concise as possible. For major architectural changes (>500 LOC excluding kernel/data/config/test), we would expect a GitHub issue (RFC) discussing the technical design and justification. Otherwise, we will tag it with rfc-required and might not go through the PR.

What to Expect for the Reviews

The goal of the vLLM team is to be a transparent reviewing machine. We would like to make the review process transparent and efficient and make sure no contributor feel confused or frustrated. However, the vLLM team is small, so we need to prioritize some PRs over others. Here is what you can expect from the review process:

• After the PR is submitted, the PR will be assigned to a reviewer. Every reviewer will pick up the PRs based on their expertise and availability.
• After the PR is assigned, the reviewer will provide status update every 2-3 days. If the PR is not reviewed within 7 days, please feel free to ping the reviewer or the vLLM team.
• After the review, the reviewer will put an action-required label on the PR if there are changes required. The contributor should address the comments and ping the reviewer to re-review the PR.
• Please respond to all comments within a reasonable time frame. If a comment isn't clear or you disagree with a suggestion, feel free to ask for clarification or discuss the suggestion.

Thank You

Finally, thank you for taking the time to read these guidelines and for your interest in contributing to vLLM. Your contributions make vLLM a great tool for everyone!

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which consists a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of default ones by unblocking the steps in your fast-check build on Buildkite UI.

Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge).

To run full CI, you can do one of these:

• Comment /ready on the PR
• Add ready label to the PR
• Enable auto-merge.

🚀

[ywang96]

Thanks for this PR! @Isotr0py

This was on my TODO list to fix and I was actually going to just import the Attention module from transformers if xformers is not present in the environment. WDYT?

[Isotr0py]

@ywang96 I totally agree that we can import the Attention implementation from transformers, since most of ViTs have implemented their SDPA and FlashAttention there. This can prevent duplicated implementation.

InternViT may be an exception for this, because they haven't added their models to transformers yet, but seems that they have had a plan for this (see here). And we may need to add a sdpa implementation for it manually.

However, one potential tricky point is that the FlashAttention2 implemented in transformers is not compatible with vllm because it requires flash_attn instead of vllm_flash_attn. And it also doesn't support tensor parallel as well. As a result, if we want to use flash attention to speed up the encoder inference in the future, we may need to re-implement it. (Please correct me if I'm misunderstanding wrong).

[Isotr0py]

@ywang96 I also add a transformers attention fallback when tensor parallel unavailable (num_heads can't be divided by tp_size).

IMO, TP on ViT shouldn't block user to use a larger tp size, especially it doesn't bring as much performance benefit as on backbone.

For example, InternViT with num_attention_heads=25 mentioned in #8055 (comment), TP on ViT is blocking user to apply tp_size=4 or higher for large model inference currently.

[DarkLight1337]

@ywang96 will be out this week so I'll work on this instead. Just a few comments.

[DarkLight1337]

LGTM now, thanks for implementing this!

[DarkLight1337]

The tests are failing. May be related to #8106?

[Isotr0py]

Ooops, I forgot to add else statement for if "vision" not in name or self.vision_tower.shard_weight:, which caused some weights not loaded correctly.

The new commit should fix this.